Attention and Information Processing

Pioneering Studies: Attention as Selection

Broadbent's Filter Model of Attention

Attention gained importance in the 1950s due to technological advancements during World War II, which presented humans with overwhelming amounts of information for extended periods. For instance, pilots had to manage numerous instruments and controls while receiving instructions, raising questions about how they prioritize and switch attention. These technology-driven questions led researchers to use tape recorders to study information intake under varying conditions (Moray, 1959).

Broadbent proposed the filter model of attention (Broadbent, 1958), as introduced in Chapter 1, Figure 1.11, to explain this. Cherry's (1953) dichotic listening experiment showed that participants could easily shadow a spoken message in the attended ear and identify the speaker's gender in the unattended ear but couldn't report the content of the unattended message. Other experiments confirmed this lack of awareness of unattended information.

Based on these findings, Broadbent (1958) developed a model to explain how we focus on one message and why other information is excluded.

The model consists of the following stages (Figure 4.3):

Sensory Memory: Holds incoming information briefly (fraction of a second) and transfers it to the filter (discussed in Chapter 5).
Filter: Identifies the attended message based on physical characteristics (e.g., voice tone, pitch, speed, accent) and allows it to pass to the detector, while filtering out other messages.
Detector: Processes attended message to determine higher-level characteristics, such as meaning. The detector processes all information that enters it.
Short-Term Memory: Holds information for 10-15 seconds and transfers it to long-term memory, which stores information indefinitely (described in Chapters 5-8).

Broadbent's Model as a Bottleneck: It restricts information flow, similar to a bottle's neck. However, unlike a bottle's neck, the filter doesn't just slow information but blocks a large portion of it, based on physical characteristics like speaking rate or voice pitch.

Early Selection Model: Broadbent's model is an early selection model because it eliminates unattended information early in the processing stream, before full analysis and meaning extraction.

Modifying Broadbent's Model: The Attenuation Model

Broadbent's filter model stimulated further research by providing testable predictions. For example, the model predicts that the meaning of unattended information shouldn't pass the filter or reach consciousness. Moray (1959) tested this with a dichotic listening experiment, instructing participants to shadow one ear's message and ignore the other.

Contrary to the early filter model, about a third of participants recognized their name when presented in the unattended ear (Wood & Cowan, 1995), known as the cocktail party effect. Gray and Wedderburn (1960) conducted a dichotic listening experiment (Figure 4.4) where the attended ear received "7 Jane," and the unattended ear received "9 Aunt 6." Participants reported hearing "Dear Aunt Jane," indicating attention shifted between ears based on meaning, challenging Broadbent's theory.

Treisman's Attenuation Model: Anne Treisman (1964) proposed a modification, the attenuation model, with selection occurring in two stages. She replaced Broadbent's filter with an attenuator (Figure 4.5).

The attenuator analyzes messages based on:

Physical characteristics (pitch, speed).
Language (syllable or word grouping).
Meaning (how word sequences form phrases).

Like Broadbent, Treisman suggested attention selects relevant information, but also considers language and meaning. The analysis proceeds only as far as necessary for identification. For similar voices, language or meaning may be needed to separate messages.

Attenuation: Unlike Broadbent's all-or-nothing filter, both attended and unattended messages pass through the attenuator, but the attended message is stronger while unattended messages are weaker – a "leaky filter" model.

Dictionary Unit: The final output is determined in the second stage, analyzed by the dictionary unit, containing words/concepts with activation thresholds (Figure 4.6). A threshold is the minimum signal strength for detection. Words with low thresholds (e.g., names, danger signals) are easily detected, even from weak signals. Uncommon or unimportant words have higher thresholds, requiring stronger signals.

Treisman proposes that the attended message gets through, along with parts of the unattended messages.

Late Selection Model

MacKay (1973) presented ambiguous sentences to the attended ear while biasing words were presented to the unattended ear and found that the meaning of the biasing word affected the participants' choice. Participants were unaware of the biasing words but these words influenced their judgments about the meaning of the ambiguous sentences.

MacKay and other theorists then developed late selection models of attention, which proposed that most of the incoming information is processed to the level of meaning before the message is selected for extended analysis and inclusion in decision processes (Deutsch & Deutsch, 1963; MacKay, 1973, Norman, 1968).

Processing Capacity and Perceptual Load

Researchers began investigating the various factors that control attention, emphasizing the primary task that participants are asked to do. Lavie (1995) considered (1) processing capacity, the amount of information people can handle, and (2) perceptual load, the difficulty of a task. Easy, well-practiced tasks have low perceptual loads, using little processing capacity. Difficult tasks are high-load tasks.

Visual tasks with distracting flankers were used in the laboratory to study these concepts.

Forster and Lavie (2008) presented displays where participants identified a target (X or N). The task was easier when the target was surrounded by the same letter and harder when surrounded by different letters.

Interestingly, a task-irrelevant stimulus slowed responses more in the easy task than the hard task, suggesting greater distraction in the easy task.

Load Theory of Attention: Lavie explains this with her load theory (Figure 4.9). With low-load tasks, remaining processing capacity handles irrelevant stimuli, slowing responses. In high-load tasks, all processing capacity is used, preventing irrelevant stimuli from being processed.

According to load theory, distraction is determined by the perceptual load of the task and the availability of attention resources. Green and Bavelier (2003) compared distraction in video game players (trained in spatial attention) and non-players, finding that high perceptual load may not reduce the effect of distracting flankers to the same extent in all participants. The compatibility effect refers to the difference in reaction time for compatible versus incompatible flankers.

Distraction and Cognitive Control

Studies have shown individual differences in distractibility in low-load tasks (Forster & Lavie, 2008). Lavie and colleagues suggest this relates to executive or cognitive control functions.

In addition to a passive early selection mechanism affected by perceptual load, there is an active late selection mechanism depending on frontal lobe functioning. Frontal lobe damage often impairs focus and the ability to ignore distractors. This cognitive control deficit is also seen in young children and older adults, possibly due to frontal lobe development and deterioration (Lavie, 2010). The frontal lobe is involved in cognitive control processes required for working memory and dual-task coordination. Working memory is the ability to temporarily keep things in mind.

Lavie et al. (2004) conducted experiments where participants kept one or six numbers in working memory (low vs. high cognitive load) while performing a flanker compatibility task. Perceptual load was low in both conditions.

The compatibility effect, or the influence of the flanker, was largest in the high working memory load condition. High cognitive load makes late selection less efficient because frontal lobe resources are needed for both keeping things in mind and using an effective selective attention filter.

The focus in attention research has shifted towards investigating the specific task and stimulus characteristics that determine how well we can keep our selective attention focus. Researchers have introduced the interplay between (1) goal-directed or top-down attention and (2) stimulus-driven or bottom-up attention.

Spatial Attention: Overt and Covert Attention

Overt attention involves shifting attention by moving the eyes, while covert attention involves shifting attention without eye movements.

Overt Attention: Scanning a Scene with Eye Movements

Scanning is necessary because good detail vision occurs only for things you are looking at directly. Central vision is the area you are looking at and falls on the fovea, which has much better detail vision than the peripheral retina, on which the rest of the scene falls. As you scanned the image, you aimed your fovea at one section after another. Each time you briefly paused on one section, you were making a fixation. When you moved your eye to observe another section, you were making a saccadic eye movement-a rapid, jerky movement from one fixation to the next.

People move their eyes about three times per second even when freely viewing a scene, creating a pattern of fixations and saccadic eye movements.

Two factors determine how people shift their attention by moving their eyes in a free-viewing situation: bottom-up, based primarily on physical characteristics of the stimulus; and top-down, based on cognitive factors such as the observer's knowledge about scenes and past experiences with specific stimuli.

Scanning Based on Stimulus Salience

Stimulus salience, such as color, contrast, or movement, can influence overt and covert attention and can cause an involuntary shift of attention, called attentional capture (Anderson, Laurent, & Yantis, 2011). Determining how saliency influences scanning involves creating a saliency map of the scene (Itti & Koch, 2000; Parkhurst, Law, & Niebur, 2002; Torralba, Oliva, Castelhano, & Henderson, 2006).

Parkhurst et al. (2002) found that the first few fixations are closely associated with the light areas on the saliency map. After the first few fixations, scanning begins to be influenced by top-down, or cognitive, processes that depend on things such as the observers' goals and expectations determined by their past experiences in observing the environment.

Researchers can determine which specific stimulus features cause a pop-out effect or an automatic shift of our attention (Wolfe & Horowitz, 2017). Stimulus salience depends on the difference with its surrounding distractors and on the homogeneity of those distractors (Duncan & Humphreys, 1989; Wolfe & Horowitz, 2017).

Scanning Based on Meaning and "Knowledge" Factors

Eye movements are determined by top-down processes associated with personal interests and scene schemas—an observer's knowledge about what is contained in typical scenes. Consequently, when Vo and Henderson (2009) showed observers pictures of a printer and a pot, they looked longer at the printer in the kitchen scene than the pot in the kitchen scene because a printer is less likely to be found in a kitchen. Additionally, Shinoda et al. (2001) found that observers are more likely to detect stop signs positioned at junctions than those positioned in the middle of a street.

Scanning Based on Task Demands

Attention is related to what we want or are asked to do. Eye movements are linked to the action the person was about to take. The person fixated on few objects or areas that were irrelevant to the task, and eye movements and fixations were closely linked to the action the person was about to take. Eye movements precede a motor action by a fraction of a second. This is an example of the "just in time" strategy—eye movements occur just before we need the information they will provide (Hayhoe & Ballard, 2005; Tatler, Hayhoe, Land, & Ballard, 2011).

Covert Attention: Directing Attention Without Eye Movements

In Posner's cueing procedure, endogenous cues always appear in the center of the computer screen (at fixation) and indicate where the participant can expect the subsequent target. This is a manipulation of top-down attention. Exogenous cues appear at one of the locations where the subsequent target could appear (left or right from fixation). This is a manipulation of bottom-up attention.

Typical results of endogenous and exogenous cueing experiments indicate that participants react more rapidly to the target when their attention was focused on the location where it was to appear, showing that information processing is more effective at the place where attention is directed.

Goal-Driven, Stimulus-Driven, and History-Driven Selection

Overt and covert attention are controlled by top-down and bottom-up factors. Attention is the result of a complex interplay between our own knowledge and goals on the one hand and the environment on the other. Attention can be automatically captured by highly salient objects in our environment.

Recently, selection history has been suggested to also play a role in the competition for attention. Selection history refers to the notion that where or what we paid attention to in the recent past can elicit a so-called lingering selection bias (Awh, Belopolsky, & Theeuwes, 2012), which is present particularly when a recently encountered stimulus has been associated with some kind of reward (Anderson, 2013).

Awh et al. (2012) proposed a framework where current goals, physical salience, and selection history feed into an integrated priority map, which someone uses to determine which object or which location should be "selected" for attention. Theeuwes (2018) suggested that the influence of a slow and effortful top-down process on visual selection is probably much smaller than what is typically assumed.

Divided Attention: Can We Attend to More Than One Thing at a Time?

Divided attention is the distribution of attention among two or more tasks. It can occur, but the ability to divide attention depends on factors including practice and task difficulty.

Divided Attention Can Be Achieved with Practise: Automatic Processing

Schneider and Shiffrin (1977) showed that participants could carry out two tasks simultaneously: (1) holding information about target stimuli in memory; and (2) paying attention to a series of "distractor" stimuli and determining whether one of the target stimuli is present among these distractor stimuli.

With approximately 900 trials of practice, participants were able to reach 90 per cent correct. Practice made it possible for participants to divide their attention to deal simultaneously with all of the target and test items. Furthermore, the many trials of practice resulted in automatic processing, which occurs (1) without intention, and (2) at a cost of none or only some of a person's cognitive resources.

Divided Attention Becomes More Difficult When Tasks Are Harder

Schneider and Shiffrin's showed that when task difficulty is increased, then automatic processing is not possible even with practice (refer to Schneider & Chein, 2003).

Distractions While Driving

Redelmeier & Tibshirani (1997) showed that the risk of a collision was four times higher when the driver was using a mobile phone than when a mobile phone was not being used. Strayer and Johnston (2001) showed that talking on a mobile phone caused participants to miss twice as many of the red lights as when they weren't talking on the phone and also increased the time it took them to apply the brakes. Strayer and Johnston concluded that talking on the phone uses cognitive resources that would otherwise be used for driving the car.

What Happens When We Don't Attend?

Not attending can cause us to miss things even if we are looking directly at them. In Traffic Psychology this is often discussed in the context of "looked-but-failed-to-see accidents," because in many cases a driver has been clearly looking in the direction of the other party but yet failed to see them (Herslund & Jorgensen, 2003; Koustani, Boloix, Van Elaslande, & Bastien, 2008).

Inattentional Blindness

Mack and Rock described experiments that showed that participants can be unaware of clearly visible stimuli if they aren't directing their attention to them. Cartwright-Finch and Lavie (2007) showed that most of the participants were "blind" to the small square, even though it was located right next to the vertical or horizontal line.

Simons and Chabris (1999) showed that nearly half of the observers failed to report that they saw the woman or the gorilla. This experiment demonstrated that when observers are attending to one sequence of events, they can fail to notice another event, even when it is right in front of them (refer to Goldstein & Fink, 1981).

Change Detection

Change blindness is the difficulty in detecting changes in scenes (Rensink, 2002). The importance of attention (or lack of it) in determining change blindness is demonstrated by the fact that when a cue was added indicating which part of a scene had been changed, participants detected the changes much more quickly (refer to Henderson & Hollingworth, 2003). The change blindness effect also occurs when the scene changes in different shots of a film.

Attention and Experiencing a Coherent World

Attention helps create binding—the process by which features such as color, form, motion, and location are combined to create our perception of a coherent object.

The binding problem has been addressed by Treisman's (1986, 1988, 1999) feature integration theory.

Feature Integration Theory

Treisman's feature integration theory says that we perceive individual features as part of the same object by doing a two-stage process, which are the preattentive stage and the focused attention stage.

Preattentive Stage

According to Treisman, the first step in processing an image of an object is the preattentive stage. In the preattentive stage, objects are analyzed into separate features. Because each of these features is processed in a separate area of the brain, they exist independently of one another at this stage of processing. Treisman and Schmidt 1982) did an experiment to show that early in the perceptual process, features may exist independently of one another.

Illusory conjunctions are combinations of features from different stimuli. According to Treisman, illusory conjunctions occur because in the preattentive stage, each feature exists independently of the others.

Focused Attention Stage

According to Treisman's model, these "free-floating" features are combined in the second stage, called the focused attention stage. During the focused attention stage, the observer's attention plays an important role in combining the features to create the perception of whole objects. To illustrate the importance of attention for combining the features, Treisman found that If repeated the illusory conjunction experiment, but participants focused all their attention on the four target items, it all the shapes were paired with their correct colours.

Balint's syndrome is a condition in in which there is an inability to focus attention on individual objects. Visual search is something we do any time we look for an object among a number of other objects. Looking for Wally in a "busy" display with lots of people and things that have red-white stripes is an example of a conjunction search. To test the idea that attention to a location is required for a conjunction search, researchers have patients conduct visual and conjunction searches.